- Joined
- Feb 7, 2007
- Messages
- 826
- Reaction score
- 23
There was a thread a while back that I cannot find after multiple searches, which talked about the statistics of vaccination and how the stats can get misused and misinterpreted to indicate large scale benefits that just aren't there. The following article made me think about that - and about how little I remember about statistics.
Can anyone recommend a basic book on healthcare stats that I can keep around as a reference, so that I can look up things like
- the difference b/w OR and RR and when it's proper to use each
- the most likely confounding factors in cohort studies
- how to interpret "intention to treat" studies, esp when a large # dropped out or were lost to follow-up
Anyway, here's the article in JournalWatch:
http://general-medicine.jwatch.org/cgi/content/full/2010/218/2
A Critical Look at Flu Vaccine Studies
The vaccine can give the impression of preventing death even when no influenza is in sight.
Although the seasonal influenza vaccine is the darling of preventive medicine these days, the data behind its efficacy in older age groups is problematic. In many observational studies, vaccination is associated with less all-cause mortality during influenza season, but bias is always a problem in this kind of study. Vaccine researchers sought to examine this bias by studying flu vaccination patterns and mortality during seasons without circulating influenza. They used a Kaiser Permanente health maintenance organization database of older adults (age, 65), which yielded almost 1.5 million person-years of follow-up during 4 years.
Overall, both the sickest individuals (as measured by a risk score that predicted future medical costs) and those most likely to die during the upcoming year were the least likely to be vaccinated. Further, patterns of vaccination were telling: For older patients (age, 75) who had been vaccinated consistently in previous years, not receiving the vaccine was associated strongly with a higher probability of death within 1 year; among those who had not been vaccinated in previous years, receiving the vaccine was associated with a higher probability of death within 1 year. Presumably, changes in individuals' vaccination habits were responses to worsening overall health.
Comment: This clever study highlights the many statistical complications of observational data: Fluctuations in overall health are all too easy to confuse with the efficacy of any given intervention. I would put this short article high on a reading list for medical students and residents, not necessarily for its message about the flu vaccine as much as for its overall wisdom about confounding factors.
Abigail Zuger, MD
Published in Journal Watch General Medicine February 18, 2010
Can anyone recommend a basic book on healthcare stats that I can keep around as a reference, so that I can look up things like
- the difference b/w OR and RR and when it's proper to use each
- the most likely confounding factors in cohort studies
- how to interpret "intention to treat" studies, esp when a large # dropped out or were lost to follow-up
Anyway, here's the article in JournalWatch:
http://general-medicine.jwatch.org/cgi/content/full/2010/218/2
A Critical Look at Flu Vaccine Studies
The vaccine can give the impression of preventing death even when no influenza is in sight.
Although the seasonal influenza vaccine is the darling of preventive medicine these days, the data behind its efficacy in older age groups is problematic. In many observational studies, vaccination is associated with less all-cause mortality during influenza season, but bias is always a problem in this kind of study. Vaccine researchers sought to examine this bias by studying flu vaccination patterns and mortality during seasons without circulating influenza. They used a Kaiser Permanente health maintenance organization database of older adults (age, 65), which yielded almost 1.5 million person-years of follow-up during 4 years.
Overall, both the sickest individuals (as measured by a risk score that predicted future medical costs) and those most likely to die during the upcoming year were the least likely to be vaccinated. Further, patterns of vaccination were telling: For older patients (age, 75) who had been vaccinated consistently in previous years, not receiving the vaccine was associated strongly with a higher probability of death within 1 year; among those who had not been vaccinated in previous years, receiving the vaccine was associated with a higher probability of death within 1 year. Presumably, changes in individuals' vaccination habits were responses to worsening overall health.
Comment: This clever study highlights the many statistical complications of observational data: Fluctuations in overall health are all too easy to confuse with the efficacy of any given intervention. I would put this short article high on a reading list for medical students and residents, not necessarily for its message about the flu vaccine as much as for its overall wisdom about confounding factors.
Abigail Zuger, MD
Published in Journal Watch General Medicine February 18, 2010
